[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RFC 3/3] block/file-posix: Let post-EOF fallocate serialize
From: |
Nir Soffer |
Subject: |
Re: [RFC 3/3] block/file-posix: Let post-EOF fallocate serialize |
Date: |
Sat, 26 Oct 2019 20:28:00 +0300 |
On Fri, Oct 25, 2019 at 1:24 PM Max Reitz <address@hidden> wrote:
>
> The XFS kernel driver has a bug that may cause data corruption for qcow2
> images as of qemu commit c8bb23cbdbe32f. We can work around it by
> treating post-EOF fallocates as serializing up until infinity (INT64_MAX
> in practice).
>
> Signed-off-by: Max Reitz <address@hidden>
> ---
> block/file-posix.c | 42 ++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 42 insertions(+)
>
> diff --git a/block/file-posix.c b/block/file-posix.c
> index 5cd54c8bff..1f5a01df70 100644
> --- a/block/file-posix.c
> +++ b/block/file-posix.c
> @@ -2713,6 +2713,48 @@ raw_do_pwrite_zeroes(BlockDriverState *bs, int64_t
> offset, int bytes,
> RawPosixAIOData acb;
> ThreadPoolFunc *handler;
>
> +#ifdef CONFIG_FALLOCATE
> + if (s->is_xfs && s->use_linux_aio &&
This limit the fix to local xfs filesystem, but the fix may be needed
to remote filesystem such
as gluster over xfs.
> + offset + bytes > bs->total_sectors * BDRV_SECTOR_SIZE)
> + {
> + BdrvTrackedRequest *req;
> + uint64_t end;
> +
> + /*
> + * The Linux XFS driver has a bug where it will discard writes
> + * submitted through the AIO interface if they happen beyond a
> + * concurrently running fallocate() that increases the file
> + * length (i.e., both the write and the fallocate() happen
> + * beyond the EOF).
> + *
> + * To work around it, we look for the tracked request for this
> + * zero write, extend it until INT64_MAX (effectively
> + * infinity), and mark it as serializing.
> + *
> + * TODO: Detect whether this has been fixed in the XFS driver.
> + */
> +
> + QLIST_FOREACH(req, &bs->tracked_requests, list) {
> + if (req->co == qemu_coroutine_self() &&
> + req->type == BDRV_TRACKED_WRITE)
> + {
> + break;
> + }
> + }
> +
> + assert(req);
> + assert(req->offset <= offset);
> + assert(req->offset + req->bytes >= offset + bytes);
> +
> + end = INT64_MAX & -(uint64_t)bs->bl.request_alignment;
> + req->bytes = end - req->offset;
> + req->overlap_bytes = req->bytes;
> +
> + bdrv_mark_request_serialising(req, bs->bl.request_alignment);
> + bdrv_wait_serialising_requests(req);
> + }
> +#endif
> +
> acb = (RawPosixAIOData) {
> .bs = bs,
> .aio_fildes = s->fd,
> --
> 2.21.0
>
>
[RFC 3/3] block/file-posix: Let post-EOF fallocate serialize, Max Reitz, 2019/10/25
- Re: [RFC 3/3] block/file-posix: Let post-EOF fallocate serialize,
Nir Soffer <=
Re: [RFC 0/3] block/file-posix: Work around XFS bug, Vladimir Sementsov-Ogievskiy, 2019/10/25
Re: [RFC 0/3] block/file-posix: Work around XFS bug, Peter Maydell, 2019/10/25